›

NormalizedScaleGradientScript and WBPP

Matthew Hughes
in PixInsight
Hi,
Apologies if this has been answered elsewhere. 
Watching Adam's new videos on WBPP. Is Local normalization in the script and then Image Integration recommended over NormalizedScaleGradient script? or just register then go to the NSGS?
If not when should you use the NSGS over Local normalisation. I have done some Broadband imaging from a Bortle 4.
Thanks
Matt

Comments

  • Matthew Hughes
    disregard i see the answer in the NSG videos published 2022-04-07
  • BlockHeadBlockHead
    Hi Matthew,

    I am curious... what did you conclude?
    I may need to make a more concrete statement on things in a more relevant way..somewhere. 

    -the Blockhead
  • Matthew Hughes
    I decided that NSG is the way to go. I just looked at the functionality of both scripts and looked at the type of data I have. I was taking a undersampled image of the SMC. Low on the horizon towards a city of 4 million from my bottle 4 skies so it’s more like 5/6.
    I had quite bad gradients and wanted to include drizzle data. I purchased the extension for NSG and compared results using WBPP and NSG.
    NSG was better hands down.
    If I didn’t have gradients like from my clubs bottle 2 skies I may just use WBPP.
    Does that make sense? Or is my logic wrong?
    Thanks
    Matt.
  • BlockHeadBlockHead
    Nothing wrong at all. All of this is relatively new and ultimately requires many people doing this to see how things go. Seems like you did your homework.

    -the Blockhead
  • Marsha Ann Wilcox
    I’ve been wondering the same thing.  So far, I have stopped WBPP at registration, moved to NSG, then back to image integration based on older videos and the new updates you’ve done recently. 

    The updated LN process in PI is interesting, but my understanding is that the NSG workflow uses an algorithm that is more precise than the LN.  I’ll admit to not having done an empirical comparison yet. 

    Thanks again for the excellent videos.
  • BlockHeadBlockHead
    Regarding the precision, John knows that I know the answer to that question.
    Whether that precision (here we really mean the photometry) matters to the average user is a different question. 
    There was a reason I was an early adopter of NSG.
    -the Blockhead 
  • Marsha Ann Wilcox
    There have been a few nice updates to NSG recently.  The new Blink section is a great addition obviating the need to exit and go to Blink to choose a reference image.  Also, the new post-process image rejection section is quite helpful. 

    I will admit to being a little confused about what it does with Image Integration settings. After NSG, Image Integration has the following settings:

    Normalization: Local Normalization
    Weights: PSF Scale SNR
    Weight Keyword: NWEIGHT (greyed out)

    I have the paid versions, so the *.xnml files are used (<n>).

    I expected Weights to be FITS keyword and NWEIGHT to be active. I've tried it both by creating a blank template on the desktop for NSG to use and letting NSG open the process itself.  Same results both ways.

    I'm wondering what I've missed.  I'd like InageInt to use the NSG weights.

    Grateful for your guidance here.

  • BlockHeadBlockHead
    Yes, a I think another update video on NSG is probably a good thing. 
    I can answer the integration question. Obviously Local Normalization needs to be set. 
    NSG no longer calculates the weight internally. That is because the weight calculation for PSF Scale SNR is identical to the formula that NSG has always used. This is the standard formula... but the input to the formula- specifically the calculation of the scaling factor, is different. PI and NSG use different methodologies for this critical parameter (the noise term is the same). 

    So NSG simply uses the engine that is already in ImageIntegration for the weight..but inputs a different scaling factor. This is why NWEIGHT is not necessary in this case.
    I think I am close with my answer. 

    I very much like the transmission graph... awesomeness!

    -the Blockhead

  • Casey Schatz
    I'll also be following this closely....I'm a big fan of NSG and Adam you saw my post on the PI forum that Recombine RGB seemed like it would (and should in my opinion) zipper the R,G,B back together to then proceed with whichever normalization method you prefer, but was surprised to learn that it assumes you are 
    in the LN workflow....
  • Marsha Ann Wilcox
    Thanks, Adam.  
    I appreciate the explanation. Very helpful indeed.




Sign In or Register to comment.